over-egging AFDDs?

BEAMA and many academics share the view that they need to look at kHz or MHz signatures

:O

Do you have a link?

Andy

The components of arc waveforms are a long way above 500Hz. Despite all that I can see how to put together an entirely compute-free AFDD. Just like early RCDs, which were little more than a differential transformer, such an AFDD might not be as thorough in its protection as a computerised one.

Yes.

Do you have a search engine (Google, Bing, DuckDuckGo or whatever)?

I expect is not getting it to trip on an arc, but getting it to not trip on the wrong sort of arc.

e.g. you don't want it tripping on switch bounce, brush arcing, or switch start lamp starting. Probably not on a circuit feeding an arc light or an arc welder.

All of which has long left me wondering how they define a standard for AFDDs /and/ a practicable means of testing manufacturer's products against it. As for end users, substitution seems the only practicable test if a faulty AFDD is suspected. Unless of course they are guaranteed to be free from defect for life - software included :)

Capacitor to harvest the hf content, rectify, current limit, big cap so it doesnt respond to half a second of arcing, magnetic trip. Simple stuff. The only load I can think of where it might be on shaky ground is the arc welder.

Arc light lol. Even if one did use one in 2022, they don't draw lots of hf current. I used to use them. No-one does now.

That's easy. The manufacturer makes them as complex & expensive as possible, determines the reuslting specs & uses that as the new standard. Anyone else's goods that don't match it fail.

I expect testers will show up. It's not to hard to simulate an arc - or even create a real one, not that anyone would now.

Have you watched any of the videos where John Ward or David Savery cobbled together rigs to generate arcs and found it remarkably difficult to get AFDDs to trip? The manufacturers seem to demonstrate them using a "magic pelicase" rig.

(and a coil to pick the hf content off)

I've seen vids of them not tripping. IIRC they were generally very slow to trip when they did. In principle they might save a huge number of house fires once they're more mature technology, but they don't seem to be there for now.

Yes, and it didn't help me.

Now could I have that link that was so easy to find please?

Andy

https://lmgtfy.app/?q=BEAMA+AFDD

and if you want more start with the the report I posted yesterday - pages 13-14

I get the impression that an arc alone is not enough to trip them. In particular they look for an arc sustained between a conductor and a carbon compound, since that reflects the nature of an arc when insulation is overheating / burning.

At which point they start to sound like voodoo

The last time I wanted to see the Gas Safety regs in the local library, I had to use a dedicated PC and could only print a maximum of 3 pages.

That was a few years ago though.

It does look as if they sometimes detect arc faults and that some (but not all) of the high-resistance faults that can cause fires involve arcing. Every fault I have come across that involved arcing was between two copper conductors, so no carbon was directly involved except as a consequence of the heat generated by the fault. The Norwegian study (which I found very interesting) also suggests that many arc-fault reports are the result of incorrect classification. I had a look at the standard (IEC62606) and found nothing to specify the expected operating frequency range. The tests themselves all involve the use of adjustable arc cells or cooked cable samples. The standard does indeed seem to involve voodoo. John

Ah, thank you. The keyword I was apparently missing is "BEAMA". Though I'm not sure how you got from there to a site in Norway.

Much to my dismay they cost 2000Kr each. That's about £170. It would be quite a bit to fit them across our whole box - and I'm not sure they'd fit.

Also to my surprise is this: "The discharge starts as an arc/spark discharge with noise frequencies (white noise) up to 300 MHz. The low and high-frequency noise eventually disappears while heat generation continues. When the fault transitions to the glowing phase, the noise frequency decreases to around 1 MHz"

Running an FFT on a 300MHz signal would take significant CPU power. They must have a hardware block to do it. The diagrams suggest this is indeed the case.

Andy

I'm not convinced. I suspect that they just look at the power in one or a few frequency bands defined with analogue filters and analyse the envelope signal and correlate it with the mains current and voltage waveforms. The standard gives some hints about that. That would seem about right for a small microcontroller such as an ARM M0.

John